In 1987, Nobel Prize-winning American economist Robert Solow famously said: “ You can see the computer age everywhere, except in productivity statistics. This observation, known as “ Solow’s paradox ”, highlighted the discrepancy between the explosion of investment in information technology and the absence of significant productivity gains in the economic data of the time.
But what did he mean, and is his remark still valid in the age of AI?
An implicit critique of technological promises
For Solow, the mass adoption of new technologies was not synonymous with an immediate improvement in productivity. IT was starting to become a mass phenomenon in business, but the expected gains were not reflected in traditional indicators.
Why was this?
Firstly, because it takes time to adapt. To take advantage of new technologies, organizations often have to rethink their processes, train their teams and adapt.
Let’s take a silly example like electricity. We often believe that it was the massive electrification of factories that led to productivity gains, but this is not entirely true. In fact, at the time, the most energy-intensive machines were located as close as possible to the nearest source of energy – often the side of the factory next to a river whose energy drove the machines. Electricity made it possible to reorganize factories by placing machines not according to their energy requirements, but rather according to their place in the production process, resulting in significant time savings and smoother production with fewer parts to move from one end of the building to the other.
But the problem with IT, which is still true today, is that businesses often rely on technology to transform them when they need to transform themselves to take advantage of it, which they do only as a very last resort (We should not expect an application to work in environments for which its assumptions are not valid).
Solow also referred to negative externalities. By this we mean that IT has introduced new complexities, such as information overload or increased overheads for system maintenance.
While it might be tempting to say that the cloud has solved the latter (or rather, shifted the problem), almost 40 years later we’re more entangled than ever in the former (Workplace Collaboration: When Technology Saturates, Productivity Stagnates and Generations Disconnect)
Finally, there’s the problem of measurement, with indicators from another era ill-suited to measuring the impact of digital technologies.
For example, the gains brought about by IT may translate into qualitative improvements that are difficult to measure, such as better collaboration, fewer errors or more personalized services, although in the end it must show up somewhere.
But traditional accounting methods tend to underestimate the impact of non-market digital activities, such as increased individual productivity or the residual effects of innovations (gradual, diffuse improvements that manifest themselves over the long term or in related activities). Generally speaking, IT creates many intangible assets that are difficult to value (Digital Economy : Towards an actual valuation of digital assets).
AI: the end or reinforcement of Solow’s paradox?
With the emergence of artificial intelligence, Solow’s paradox is back in the spotlight. AI promises enormous gains in efficiency, but its direct impact on productivity remains hard to pin down, at least in the short term (Generative AI: what impact on work performance?[FR]).
And the problems are similar to those experienced at the start of the computer age.
In terms of integration time, AI requires specific skills, costly infrastructures and major organizational adjustments, the scale of which we are only just beginning to perceive.
Secondly, AI amplifies the disparities between businesses capable of adopting it effectively and others, which can create a zero-sum game: what will be gained somewhere will be destroyed elsewhere.
Finally, and not surprisingly, traditional indicators such as GDP or hourly productivity struggle to reflect qualitative gains or indirect effects, such as improved decision-making or personalized services (Productivity: what if quality was the new quantity?).
However, AI raises new questions, such as massive automation (AI’s substitution of many human functions could affect global demand or provoke societal transformations that are difficult to anticipate) and the concentration of profits, accentuating a phenomenon of wealth polarization on the one hand, and ultra-pauperization on the other (Towards a golden age of welfare and precariousness?).
But here we are dealing with major societal issues which, if we face them unprepared and unanticipated (as we are on the way to doing), will make our productivity worries seem minor (The challenges posed by AI are not technological, but must be met today.).
The limits of technological optimism
Whether we’re talking about IT or AI, we come back to the well-known limits of technological solutionism (To solve anything, click here). It would be tempting to convince ourselves that AI or any other technology will single-handedly solve all the problems of business or even society, but this is precisely where Solow’s paradox comes into its own.
First of all, and once again, technology alone is not enough. The impact of tools depends on their integration into appropriate organizational and societal models. Suffice to say, we’re nowhere near there, and the vacuity of public debate on subjects like AGI is despairing.
Then there are the negative externalities. AI, like past technologies, can and certainly will create new forms of complexity through phenomena such as algorithmic bias or loss of control over critical systems.
Then there’s the illusion of productivity. As I said in a previous article, it’s vital to redefine what we mean by “productivity”. Is it to produce more with less? Improving quality of life? Reduce our environmental impact? Delivering more quality?
Bottom line
Solow’s paradox may sound like something out of the 80s, but it remains a useful prism for understanding the dynamics of the economy in the face of technology.
Technological promises must be listened to with nuance and caution, recognizing both their transformative potential and their limits. Ultimately, it’s not the technologies that change the world, but the way we choose to use them.
Image: Robert Solow by Olaf Storbeck.jpg
